Can you raed this? Olny 55 plepoe out of 100 can.I cdnuolt blveiee that I cluod aulaclty uesdnatnrd what I was rdanieg. The phaonmneal pweor of the hmuan mnid, aoccdrnig to a rescheearch at Cmabrigde Uinervtisy, it dseno't mtaetr in what oerdr the ltteres in a word are, the olny iproamtnt tihng is that the frsit and last ltteer be in the rghit pclae. I awlyas tghuhot slpeling was ipmorantt!
Now, a computer would have a very hard time with this, but as a native English speaker I read this with ease. As Marvin Minsky has pointed out (no relation to Hyman), computers still can't tell the difference between a dog and a cat, something my 3 year old daughter does with ease. The theme of Godel, Escher, Bach by Doug Hofstadter is that the mind is insanely good at finding patterns, and what is easy for us can be really difficult logically, and what is hard for us is easy for computers.
When I'm peevish, I'm amazed how how stupid most people are, how little original thought is created and appreciated, how faux rigor from false assumptions does not overcome them, how many phonies there are. But then, I remember stuff like this, and think that most us are quite clever.
5 comments:
Computer can actually do fairly well at telling dogs and cats apart: http://www.parc.com/content/attachments/machine-learning-attacks.pdf describes a classifier with 80% accuracy in classifying dogs vs. cats. Within 5 years, machine learning should be able to achieve near-human classification accuracy at a significantly higher speed.
I find it more bizarre that 45% of people tested can't read it. I wonder why it's not a universal ability, like the ability to distinguish cats and dogs is. Is it correlated with IQ or some other measure of cognitive ability?
http://en.wikipedia.org/wiki/Typoglycemia
To tell a dog from a cat, pull the tail. If there's no tail it's a Manx.
Humans are just great big frauds. Experts at passing tests and utilizing universe reduction to help them read.
Or should I say:
Hmns r jst grt bg frds. xprts t pssng tsts nd tlzng nvrs rdctn t hlp thm rd.
That one is much harder, even with the plaintext available.
As AI gets pushed in that same direction (applying contextual dictionaries and grammars to reduce the search space), they are already becoming much better at many "hard" task such as species recognition and speech to text, and I think the prediction by the anonymous robot above might be only slightly optimistic.
Until recently, spelling wasn't standardized. People winged it, just like the quoted text, but seemed to communicate just fine.
Post a Comment